非线性模型预测控制及其在发电过程控制中的应用

发布时间：2018-04-14 01:21

本文选题：非线性 + 约束　；参考：《华北电力大学》2014年博士论文

【摘要】：工业过程对象由于操作点在整个操作区域范围内频繁变化,往往呈现非线性特性。实际工业过程中还存在大量的物理约束。因此,解约束非线性优化控制问题是工业过程面临的一个巨大挑战。模型预测控制是一种基于模型的先进优化控制策略,它采用滚动时域方式,并且直接处理过程约束。传统的模型预测控制策略往往只针对线性系统有效。一般情况下,非线性模型预测控制采用序列二次规划、罚函数法、KKT条件等方法来在线解优化问题。对大部分复杂系统来说,得到的非线性优化问题往往是非凸的,且在线计算负担很大。本文研究了非线性模型预测控制优化方法和其在复杂系统中的应用,主要工作概括如下：考虑一种仿射型非线性状态空间模型,提出了一种基于输入输出反馈线性化的非线性模型预测控制策略。构造了可以保证算法收敛性的迭代二次规划方法来处理约束。该算法可以在整个预测时域上得到可行解。在连续系统数字例子和连续搅拌反应器的例子中证明了其有效性。所构造的基于输入输出反馈线性化的预测控制策略进而推广应用到风力发电系统的双馈电机控制和永磁同步电机控制上。这两种快过程都具有很强的非线性,也包含很多不确定性因素。双馈电机的动态性能和风速相关,其转矩是定子和转子电流的非线性函数。永磁电机内部具有非线性耦合,以及不可测的扰动,而且参数变化,负载也会频繁变化。由于双馈电机模型中总相对阶小于状态变量个数,双馈电机只能进行输入输出反馈线性化。而永磁电机模型中总相对阶等于状态变量个数,永磁电机可以进行精确的状态反馈线性化。和现有的控制方法相比,这种策略在保证算法收敛性的同时能够降低在线计算负担,保证控制系统的实时性。超超临界机组是一种高效、高燃料利用率、低排放的先进发电设备。由于系统规模庞大、大范围变工况下具有强非线性等特性,常规控制难以实现在负荷跟踪和电网频率扰动情况下的快速稳定响应的协调控制。本文构造了包含经济指标优化层和跟踪性能优化层的多层控制结构,在此多层控制结构基础上构造了超超临界机组的分级非线性模型预测协调控制。在此结构中,上层基于超超临界机组的非线性模型,对包含经济指标和环境指标的目标函数进行优化获得下层机组运行的给定值。上层的目标是降低机组的运行成本,抑制机组本身和电网产生的扰动。下层实现超超临界机组对上层产生的优化给定值的精确跟踪。因为分级非线性模型预测协调控制的优化问题是非凸问题,对超超临界机组采用模糊神经网络建模使其转为凸二次规划。针对负荷跟踪和电网频率扰动问题进行了详细分析,仿真证明了所提出的分级非线性模型预测协调控制策略的有效性。
[Abstract]:Because the operating point of the industrial process object changes frequently in the whole operation area, it often presents nonlinear characteristics.There are still a lot of physical constraints in the actual industrial process.Therefore, solving constrained nonlinear optimal control problems is a great challenge for industrial processes.Model Predictive Control (MPC) is an advanced optimal control strategy based on model. It adopts rolling time domain and directly deals with process constraints.The traditional model predictive control strategy is only effective for linear systems.In general, nonlinear model predictive control uses sequential quadratic programming, penalty function method and KKT condition to solve on-line optimization problems.For most complex systems, the obtained nonlinear optimization problems are often non-convex and the online-computing burden is very large.In this paper, the nonlinear model predictive control optimization method and its application in complex systems are studied. The main work is summarized as follows: an affine nonlinear state space model is considered.A nonlinear model predictive control strategy based on input and output feedback linearization is proposed.An iterative quadratic programming method, which can guarantee the convergence of the algorithm, is constructed to deal with constraints.This algorithm can obtain feasible solution in the whole prediction time domain.The effectiveness of the method is proved by numerical examples of continuous systems and examples of continuous stirred reactors.The proposed predictive control strategy based on input and output feedback linearization is extended to wind power system doubly-fed motor control and permanent magnet synchronous motor control.Both of these fast processes have strong nonlinearity and many uncertainties.The dynamic performance of the doubly-fed machine is related to the wind speed, and its torque is a nonlinear function of stator and rotor current.The permanent magnet motor has nonlinear coupling and undetectable disturbance, and the load will change frequently when the parameters change.Because the total relative order in the model is less than the number of state variables, the doubly-fed machine can only linearize the input and output feedback.The total relative order in the permanent magnet motor model is equal to the number of state variables, and the permanent magnet motor can be linearized precisely by state feedback.Compared with the existing control methods, this strategy can not only guarantee the convergence of the algorithm, but also reduce the burden of on-line computation and ensure the real-time performance of the control system.Ultra-supercritical unit is an advanced generation equipment with high efficiency, high fuel efficiency and low emission.Due to the large scale of the system and the strong nonlinearity in a large range of variable operating conditions, it is difficult for the conventional control to realize the coordinated control of the fast and stable response under the condition of load tracking and frequency disturbance of the power network.In this paper, a multi-layer control structure including economic index optimization layer and tracking performance optimization layer is constructed. Based on the multi-layer control structure, the hierarchical nonlinear model predictive coordination control for ultra-supercritical units is constructed.In this structure, the upper layer optimizes the objective function including economic index and environment index based on the nonlinear model of ultra-supercritical unit to obtain the given operating value of the lower unit.The goal of the upper layer is to reduce the operating cost of the unit and restrain the disturbance caused by the unit itself and the power grid.The lower layer realizes the accurate tracking of the optimal given value generated by the upper layer of the ultra supercritical unit.Because the optimization problem of hierarchical nonlinear model predictive coordination control is a non-convex problem, fuzzy neural network is used to model the ultra-supercritical unit to transform it into convex quadratic programming.The load tracking and frequency disturbance problems are analyzed in detail. The simulation results show the effectiveness of the proposed hierarchical nonlinear model predictive coordinated control strategy.
【学位授予单位】：华北电力大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TM614;O221.2

【参考文献】